FAMO: Fast Adaptive Multitask Optimization

TL;DR

FAMO is a novel multitask optimization method that adaptively balances task losses efficiently using constant space and time, outperforming existing gradient manipulation techniques in large-scale scenarios.

Contribution

Introduces FAMO, a dynamic weighting approach for multitask learning that reduces task losses efficiently with constant resource requirements.

Findings

FAMO achieves comparable or better performance than state-of-the-art methods.

FAMO significantly reduces computational and memory overhead.

Experimental results cover supervised and reinforcement learning tasks.

Abstract

One of the grand enduring goals of AI is to create generalist agents that can learn multiple different tasks from diverse data via multitask learning (MTL). However, in practice, applying gradient descent (GD) on the average loss across all tasks may yield poor multitask performance due to severe under-optimization of certain tasks. Previous approaches that manipulate task gradients for a more balanced loss decrease require storing and computing all task gradients ($\mathcal{O}(k)$ space and time where $k$ is the number of tasks), limiting their use in large-scale scenarios. In this work, we introduce Fast Adaptive Multitask Optimization FAMO, a dynamic weighting method that decreases task losses in a balanced way using $\mathcal{O}(1)$ space and time. We conduct an extensive set of experiments covering multi-task supervised and reinforcement learning problems. Our results indicate that FAMO achieves comparable or superior performance to state-of-the-art gradient manipulation techniques while offering significant improvements in space and computational efficiency. Code is available at \url{https://github.com/Cranial-XIX/FAMO}.